proomt

Search

Search posts, papers, and topics

All posts

Hugging Face Daily PapersBin Lin, Bo Zhao, Boyang Wang2 min readpaperadvanced

StepAudio 3 Gen Technical Report

Summary

StepAudio 3 Gen is a general‑purpose audio generation model that replaces diffusion with a discrete autoregressive generator over residual vector quantization tokens. Using a 16‑layer RVQ tokenizer and progressive pretraining, it reaches state‑of‑the‑art zero‑shot TTS and voice‑design performance while handling speech, vocals, sound effects, and music.

  • Uses a discrete autoregressive generator over RVQ tokens instead of diffusion for audio synthesis.
  • Tokenizer encodes audio at 12.5 Hz into a shared 16 × 2048 residual code space, preserving semantic and waveform information.
  • First codebook is predicted autoregressively over time; a lightweight causal Transformer fills the remaining 15 codebooks across the codebook axis.
  • Interference‑aware progressive pretraining lets the model acquire audio capabilities while retaining the LLM's textual abilities.

Engineers building multi‑modal generative systems need a unified, token‑based approach that scales across audio domains and integrates with large language models.

8/10

Related reading

  1. StepAudio 3 Realtime Technical Report

    StepAudio 3 Realtime is an audio‑language foundation model that runs a continuous listen‑converse‑think‑act loop. It introduces Deep Perception for rich acoustic cue extraction, Seamless Duplex for handling pauses/back‑channels, and a Think‑While‑Speaking mechanism that lets the model reason in parallel with speech output. On benchmarks it scores 73.0 macro avg on StepAudioChat, 90.6 on MMSU, 98.…

    Hugging Face Daily Papersarxiv.org2 minpaper
  2. Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech

    Researchers train a compact 82 M‑parameter Thai fixed‑voice TTS model using synthetic speech generated by a large voice‑cloning teacher, requiring only a 15‑second real reference. The student achieves 68.2% keyword accuracy and 91.4% pause precision, outperforming its teacher on pause placement and enabling on‑device inference.

    Hugging Face Daily Papersarxiv.org1 minpaper
  3. webaudio fingerprinting on alibaba

    Firefox made WebAudio output deterministic in v118, collapsing most variance to three CPU‑dependent buckets (x86 without FMA, x86/x64 with FMA, ARM NEON). Telemetry shows 99.24% of users fall into these buckets; a long tail of 23 distinct values affects <0.1% of users. Ongoing bugs aim to merge the remaining buckets. Alibaba’s fingerprint script was extracted (two SHA‑256 hashes) and shown to be…

    Mozilla Automation Teamritter.vg3 min
  4. PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control

    PhysStream introduces a two‑stage autoregressive video generator that uses online‑derived positional and tracking maps (structured scene memory) and sparse velocity‑increment signals to enable fine‑grained, physics‑grounded control of multi‑object tabletop scenes. It cuts motion distribution error by 33 % and trajectory error by 12 % versus strong baselines, and wins 85 % of human preference test…

    Hugging Face Daily Papersarxiv.org1 minpaper
  5. FRAUDSkill: Structured Frozen-Weight Skill Optimization for Audio Anti-Fraud Detection

    FRAUDSkill is a framework that keeps a pretrained audio‑language model frozen and learns an external layer of skill programs, routing policies, and decision rules to meet a structured anti‑fraud detection protocol. On the TeleAntiFraud benchmark it reaches 73.5% Macro‑F1 (≈32% improvement) while cutting invalid predictions to 1.94%.

    Hugging Face Daily Papersarxiv.org1 minpaper